WHAT NEW IN X11.04 WED

AI collections from X · 22 collections · 67,905 posts

jevIMAGE

Invoice pipeline: Jev sorts inbox perfectly, but a 4B fine-tune wins on payment decisions

Jev classified the inbox 200/200 with no training needed, but on deciding whether to pay, a 4B fine-tuned model scored 0.98 against Jev's 0.79.

Jacek Golebiowski
@j_golebiowskiArchived 2026.09.24Originally posted 2026.09.23
Invoice pipeline: Jev sorts inbox perfectly, but a 4B fine-tune wins on payment decisions
Jacek Golebiowski @j_golebiowskiOriginally posted 2026.09.23
Open original on X

Where does Jev end and a fine-tuned small model begin? We built an invoice pipeline with both; sorting the inbox: Jev 200/200, nothing to train. Deciding whether to pay: Jev lands at 0.79 but a 4B fine-tuned to reason first gets 0.98. Details from @distil_labs below.

x.com/j_golebiowski/status/2102782627442819504Archived 2026.09.24 10:15